Purpose. Some criminal trials turn on evaluations of credibility of the complainant and the accused. When credibility is based on how a witness testifies, an evaluation of one party should not influence an evaluation of the other party. If credibility evaluations are bidirectional, fundamental principles of criminal law may be offended.

Methods. Six hundred and thirty seven undergraduates read a vignette that described a sexual assault (SA) or a motor vehicle accident (MVA). Karen, the complainant in the SA case and bystander witness in the MVA case, was described as 5, 13, or 20 years old. The vignette was a summary of the police investigation and, in three conditions, the trial. Trial information was manipulated in one of three ways: no information concerning how Karen testified, Karen's testimony was described positively (pro-prosecution), or described negatively (anti-prosecution). Participants then rated the perceived credibility of Karen and the accused (Bob) and the probability that Bob was guilty.

Results. Karen was viewed more positively in the pro-prosecution condition and more negatively in the anti-prosecution. When Karen was judged to be less credible, Bob was rated as more credible and less likely to be guilty. When Karen was seen as more credible, Bob was viewed as more likely to be guilty.

Conclusions. This bidirectional effect that the manner in which the prosecution witness testified affected perceptions of the accused and probability of guilt, in certain circumstances, may compromise fundamental principles of criminal law and be a reversible error. We offer possible solutions for future empirical testing.